Fix convergence issues and switch to LLaMA in the SST-2 example #343
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR switches the model to LLaMA in the SST-2 example notebook, as well as fixes NaNs appearing during training by using proper mixed precision with loss scaling. I've removed deep p-tuning for now for the sake of simplicity, will restore and fix the PersonaChat notebook once I fix it as well.
The example run is here: https://wandb.ai/mryab/bloom-sst-2/runs/3e0suayd?workspace=user-mryab. As we can see, the loss is still a bit unstable, bug at least we have no outright divergence:
When merged, this should fix #327, #283 and #241.